mantle compositional control on the extent of mantle...

Mantle compositional control on the extent of mantle melting,crust production, gravity anomaly, ridge morphology,

and ridge segmentation:a case study at the Mid-Atlantic Ridge 33^35³N

Yaoling Niu a;b;*, Daniel Bideau c, Roger Hekinian c, Rodey Batiza d

a Department of Earth Sciences, Cardi¡ University, P.O. Box 914, Park Place, Cardi¡ CF10 3YE, UKb ACQUIRE, The University of Queensland, Brisbane, Qld. 4072, Australiac IFREMER, Centre de Brest, DRO/GM, B.P. 37, 29287 Plouzane, France

d National Science Foundation, Arlington, VA 22230, USA

Received 3 March 2000; received in revised form 17 January 2001; accepted 24 January 2001

Abstract

Mantle temperature variation and plate spreading rate variation have been considered to be the two fundamentalvariables that determine the extent of mantle melting and ocean crust production. Along the length of a V200 kmportion of the Mid-Atlantic Ridge (MAR) between the Oceanographer (35³N) and Hayes (33³N) transforms, themantle potential temperature is the same, the plate spreading rate is the same, but the extent of mantle melting andcrustal production vary drastically. In addition to the typical crustal thickness variation on ridge segment scales at theMAR, i.e. thicker at segment centers and thinner at segment ends, there exist between-segment differences. Forexample, the V90 km long segment OH-1 is magmatically robust with a central topographic high, thick crust, and alarge negative gravity anomaly whereas the V45 km long segment OH-3 is magmatically starved with a deep rift valley,thin crust and a weak negative gravity anomaly. We demonstrate that the observed differences in the extent of mantlemelting, melt production and crustal mass between segments OH-1 and OH-3 are ultimately controlled by their fertilemantle source compositional difference as reflected by the lava compositional differences between the two segments:s 70% of OH-1 samples studied (N = 57) are enriched MORB with [La/Sm]N s 1, but s 85% of OH-3 samples studied(N = 42) are depleted MORB with [La/Sm]N 6 1. Calculations show that the mean OH-1 source is more enriched inincompatible elements, total alkalis (V0.36 wt% Na2O and V0.09% K2O) and H2O content (V280 ppm) than themean OH-3 source, which is depleted of incompatible elements, total alkalis (6 0.17% Na2O and 6 0.01% K2O) andH2O content (V70 ppm). These fertile compositional differences result in significantly reduced solidus temperature ofOH-1 source over that of OH-3 source, and allows melting to begin at a significantly greater depth beneath OH-1 (V90km) than beneath OH-3 (6 60 km), leading to a taller melting column, higher degrees of decompression melting, greatermelt production, thus thicker crust and more negative gravity anomaly at OH-1 than at OH-3. We emphasize thatfertile mantle source compositional variation is as important as mantle temperature variation and plate spreading ratevariation in governing the extent of mantle melting, crustal production, and MORB chemistry. The buoyancy-driven

0012-821X / 01 / $ ^ see front matter ß 2001 Elsevier Science B.V. All rights reserved.PII: S 0 0 1 2 - 8 2 1 X ( 0 1 ) 0 0 2 5 5 - 2

* Corresponding author. Tel. : +44-29-2087-6411; Fax: +44-29-2087-4326; E-mail: niuy@cardi¡.ac.uk

EPSL 5771 22-3-01

Earth and Planetary Science Letters 186 (2001) 383^399

www.elsevier.com/locate/epsl

focused mantle upwelling model better explains the observations than the subcrustal melt migration model. Futuremantle flow models that consider the effect of fertile mantle compositional variation are expected to succeed inproducing along-axis wavelengths of buoyant flow comparable to the observed size of ridge segments at the MAR. Wepropose that the size and fertility of the enriched mantle heterogeneities may actually control the initiation andevolution of ridge segments bounded by non-rigid discontinuities at slow-spreading ridges. ß 2001 Elsevier ScienceB.V. All rights reserved.

Keywords: Mid-Atlantic Ridge; mid-ocean ridge basalts; mantle; ridge segmentation; crust; ocean £oor; gravity anomalies

1. Introduction

The ocean crust is the net product of magmatic,hydrothermal and tectonic processes operatingwithin and below the global ocean ridge system.In contrast to the relative uniformity along thefast-spreading East Paci¢c Rise (EPR), crustal ac-cretion at slow-spreading ridges, such as the Mid-Atlantic Ridge (MAR), is complex. The MAR ishighly segmented to various lengths by both per-manent transforms and non-rigid discontinuities,and has large variations in both across-axis mor-phology and along-axis topography [1,2]. ManyMAR segments are characterized by a bathymet-ric high in the segment center, forming a three-dimensional (3-D) bathymetric structure on seg-ment scales. Correlated with these bathymetricstructures are circular gravity lows or `bull'seyes' [3^5] interpreted to re£ect an increase incrustal thickness from segment ends to the seg-ment center by up to 4 km [3^7]. What is partic-ularly important is the signi¢cant correlation ofridge segment length with the amplitude of thegravity anomaly and the magnitude of the in-ferred crustal thickness variation [3^5]. Long ridgesegments are typically associated with large man-tle Bouguer gravity lows, shallow hour-glass-shaped rift valleys that widen and deepen towardssegment ends, and have greater mean crustalthickness and larger along-axis crustal variation.In contrast, short segments are characterized bysmaller gravity lows, deeper and low-relief riftvalleys, and have smaller along-axis crustal varia-tion and thinner crust [3,5,8].

While these correlated variations are well-docu-mented and point to a genetic link among theseparameters, the nature of such a link remains un-clear. The actual cause of the contrasting topo-

graphy, morphology, gravity anomaly and crustalthickness between long and short ridge segments,particularly when they are in spatial proximity, isunknown. In this paper, we present geochemicaldata on densely sampled lavas from the two con-trasting ridge segments (OH-1 and OH-3 in Fig.1) between the Oceanographer and Hayes trans-forms along the MAR at 33^35³N. The data sug-gest that the fertile mantle compositional di¡er-ence between OH-1 and OH-3 controls theobserved di¡erences between the two segments.Mantle source heterogeneity beneath ocean ridgesis a well-known fact. Recent studies have in par-ticular emphasized the mantle source (vs. process-es) control on the observed MORB chemistry [9^12] and appealed to a discriminate use of MORBmajor element composition in estimating meltingconditions [12,13]. However, the consequentialrole of fertile mantle compositional variation indetermining the extent and depth of mantle melt-ing and ocean crust production remains largelyoverlooked.

2. MAR at 33^35³N and samples

The MAR between the Hayes and Oceanogra-pher transforms is one of the best-studied seg-ments of the MAR. Reconnaissance studies alongthe North MAR by Schilling and co-workers[14,15] revealed the ¢rst-order physical and chem-ical consequences of plume^ridge interactions.The 1991 FARA-Sigma expedition [5,16] pro-duced detailed bathymetric and gravity maps forthe MAR extending from south of the Hayestransform at 33³N to the northern edge of theAzores Platform near 40³N. The 1992 FARA-Fa-zar expedition [17,18] allowed detailed sampling

EPSL 5771 22-3-01

Y. Niu et al. / Earth and Planetary Science Letters 186 (2001) 383^399384

along this same portion of the MAR. These earlystudies have targeted the Oceanographer^Hayesarea for detailed investigations [7,8,19^26] becausethis area consists of contrasting ridge segmentswith distinctive topography, morphology andgravity anomalies, hinting at clues about crustalaccretion processes at slow-spreading ridges.

The northern segment OH-1 is V90 km long,and has a central topographic high with a medianridge (1700^2000 m), a chain of o¡-axis volcanoestrending at a high angle across the ridge axis, alarge (V40 mGal peak-to-trough) negative man-tle Bouguer anomaly (MBA), and a thick crust(Fig. 1). This segment is clearly magmatically ro-bust [19^22] with possibly a magma chamber/lenspresent at its midpoint [23]. In contrast, segmentsOH-2 and OH-3 to the south are shorter (V59and 45 km respectively), have deep (3000^3500 m)U-shaped axial valleys, and much smaller (V25and 6 20 mGal respectively) MBAs. In situ sub-

mersible observations and exposures of mantleperidotites at the non-transform o¡set betweenOH-2 and OH-3 con¢rm the presence of anoma-lously thin crust due to starved magmatism [19^22] at these segments. The contrasting di¡erencein melt generation and crust production on such asmall (V200 km long Hayes^Oceanographer seg-ment) geographic scale presents a challenge to ourcurrent theories. Mantle temperature variation[27^29] and plate spreading rate variation [13,30]are considered to be the primary controls on theextent of mantle melting and crustal production.As the plate spreading rate is essentially the sameat these adjacent ridge segments, di¡erences inmantle potential temperature would be invoked.Is the mantle hotter beneath OH-1 than beneathOH-2 and OH-3? Is it possible that the Azoreshotspot to the north has greater thermal in£uenceon OH-1 than on the more southerly OH-2 andOH-3? Is there any alternative mechanism that

Fig. 1. a: Portion of the North MAR and the vicinity showing the study area between the Hayes and Oceanographer transformsduring the OCEANAUT expedition after [19,20]. b and c: Simpli¢ed maps of the bathymetry and MBAs of the area. d: Along-axis pro¢les of ridge axial depth (bottom solid line), MBA (top solid line), and the interpreted thermal e¡ect on the gravityanomalies. b, c and d are after [5].

EPSL 5771 22-3-01

Y. Niu et al. / Earth and Planetary Science Letters 186 (2001) 383^399 385

may have led to the contrasting di¡erences inmantle melting and crustal production? Geo-chemistry of basalts, the very product of mantleand crustal processes, must carry the messageabout the nature of these processes.

The 1995 OCEANAUT expedition aboard R/VNadir with the submersible Nautile made signi¢-cant in situ observations, and collected V300rock samples from magmatically robust OH-1and starved OH-3 [19^22,26]. We discuss herethe whole-rock major element data of 27 repre-sentative samples and whole-rock trace elementdata of 99 representative samples from the tworidge segments. Most samples studied are rela-tively fresh aphyric basalts. Some are freshglasses, and others are plagioclase-phyric or pla-gioclase-olivine phyric basalts, for which only ma-trix portions were analyzed to avoid the e¡ect ofcrystal accumulation. Some o¡-axis samples areapparently weathered, in which case mobile ele-ments such as Rb, Cs, Pb, Li are apparently af-fected, and thus are not discussed.

3. Data and interpretations

Major elements on 13 OH-1 samples and 14OH-3 samples were analyzed by ICP-AES at Uni-versite de Bretagne Occidentale [31]. Trace ele-ments on 57 OH-1 samples and 42 OH-3 sampleswere analyzed by ICP-MS at the University ofQueensland [32]. The complete analyses and sam-ple descriptions are available on an EPSL OnlineBackground Dataset1. Figs. 2^4 show the datagraphically.

Fig. 2 shows variations of highly incompatibleelements like Ba and Th and ratios of more-over-less incompatible elements such as Nb/Zr, [La/Sm]N, [Sm/Yb]N and Zr/Y as a function of MgOand Na2O for lavas from both OH-1 and OH-3.Lavas from OH-1 and OH-3 form distinctivegroups. All but one OH-1 samples have lowerMgO than OH-3 samples. OH-1 samples haveon average signi¢cantly higher abundances and

ratios of incompatible elements, including Na2Othan OH-3 samples. The low MgO and highNa2O of the OH-1 samples could be interpretedsolely as resulting from greater extents of frac-

Fig. 2. Variations of incompatible element abundances (Baand Th in ppm) and ratios (Nb/Zr, [La/Sm]N, [Sm/Yb]N andZr/Y; where the subscript N denotes chondrite-normalizedvalues) as a function of MgO and Na2O (wt%) for lavasfrom both ridge segments OH-1 and OH-3. The shadedbands indicate trends of low-pressure fractionation (with de-creasing MgO and increasing Na2O) on these plots derivedform isotopically well-constrained N-type MORB from thenorthern EPR region [12,32,57].

1 http://www.elsevier.nl/locate/epsl; mirror site: http://www.elsevier.com/locate/epsl

EPSL 5771 22-3-01


tionation than OH-3 samples, so could be the in-verse correlations of MgO and positive correla-tions of Na2O with the abundances and ratiosof incompatible elements on these plots. However,neither the abundance levels nor detailed slopesde¢ned by either OH-1 or OH-3 lavas can be ex-plained by low-pressure fractionation as the latterwould produce much shallower slopes as indi-cated by the shaded bands. High-pressure frac-tionation and the e¡ect of varying volatile con-tents in primary melts would produce liquidlines of descent that di¡er from the low-pressure

fractionation trends, but these would not causesigni¢cant fractionation in Nb/Zr, La/Sm, Sm/Yb and Zr/Y ratios. Therefore, the di¡erent abun-dance levels and incompatible element ratios de-¢ned by OH-1 and OH-3 lavas on these plots arenot caused by fractionation, but inherited fromtheir primary magmas.

If we assumed the abundances and ratios ofthese incompatible elements, including Na2O, re-sulted from varying extents of melting only, thenwe cannot avoid the conclusion that OH-1 lavaswere produced by very low extents of melting,whereas OH-3 lavas by very high extents of melt-ing. This conclusion is undoubtedly wrong as itcontradicts the observations! In fact, the largevariation in incompatible element ratios in Fig.2 can only be explained by melting composition-ally heterogeneous sources [12^15,32,33].

Fig. 3 plots along-axis variation of [La/Sm]Nand the abundances of representative incompati-ble elements like Th and Zr at OH-1 and OH-3.Highly depleted lavas with extremely low Th, Zrand [La/Sm]N exist at both OH-1 and OH-3, butlavas from OH-1 are on average signi¢cantlymore enriched with many unusually enrichedMORB lavas having [La/Sm]N s 1.5 and up to3.0, indicating the presence of an enriched andheterogeneous source beneath OH-1 (vs. OH-3).Note that the along-£owline o¡-axis samples(open circles) across central topographic high ofOH-1 display a similar range to the axial lavas,suggesting that the large compositional range and`spike' in the segment center de¢ned by `zero-age'lavas have been persistent in the last v2.5 Ma.Also note that less depleted or enriched lavaspreferentially occupy the center of a segment atboth OH-1 and OH-3 with the maxima coincidingwith the along-axis depth pro¢le minima (bottompanels) and MBA maxima (Fig. 1b and d). Thisphenomenon has been observed elsewhere alongthe MAR [17,18,33].

Plots of incompatible element abundances (e.g.Ba, U, Ta and Sr) and ratios of more-over-lessincompatible elements (e.g. Nb/La, La/Sm, Zr/Yand Hf/Yb) as a function of axial depth revealscattered yet signi¢cant correlations (Fig. 4). De-pleted lavas dominate at deep portions of theridge at both OH-1 and OH-3, but the highly

Fig. 3. Along-axis variation of Th and Zr abundances (ppm)and [La/Sm]N ratio to show that (1) large lava compositionalvariation exists on small scales at both OH-1 and OH-3; (2)highly depleted lavas with extremely low Th, Zr and [La/Sm]N exist at both OH-1 and OH-3, but lavas from OH-1are on average signi¢cantly more enriched with many unusu-ally enriched E-type MORB lavas ([La/Sm]N s 1.5); and (3)less depleted or enriched lavas prefer to occupy the center ofa segment in both cases with the maxima coinciding with thealong-axis depth pro¢le minima (bottom panels) and MBAmaxima (Fig. 1b and d). Open symbols on OH-1 panels areo¡-axis samples. The shaded areas on the bottom panels aresampling depths.

EPSL 5771 22-3-01


enriched ones are exclusively associated withtopographic high at OH-1. This observation re-£ects a genetic link between the topographichigh and the enriched lavas that make up thehigh. Histograms (not shown) indicate that morethan 70% OH-1 samples have [La/Sm]N s 1whereas more than 85% OH-3 samples have [La/Sm]N 6 1 (see EPSL Online Background Data-set1).

4. Discussion

The data and foregoing discussion raise two¢rst-order questions: (1) Why are highly enrichedlavas exclusive to magmatically robust OH-1, butabsent at magmatically starved OH-3? (2) Whydo enriched or less depleted lavas preferentially

occupy the segment center? The questions canbe answered by analyzing the contrasting di¡er-ences between OH-1 and OH-3 in terms of funda-mental variables that govern the magmatic accre-tion at ocean ridges.

4.1. The cause of the contrasting di¡erencesbetween OH-1 and OH-3

The in£ated topography and morphology andlarge MBA at OH-1 (Fig. 1) and submersible ob-servations indicate unequivocally that this seg-ment is magmatically robust with a large crustalmass relative to OH-3, which is magmaticallystarved with a small crustal mass. Greater crustalmass results from a greater amount of mantlemelt produced and extracted. Assuming completemelt extraction, then the crustal mass (Mc) pro-duced per unit time is the product of the extent ofmelting (F) from a given parcel of fertile mantleand the amount of the fertile mantle melted (Mf )per unit time. That is, Mc = FUMf . As mantlemelting beneath ridges results from decompres-sion, Mf would be a function of mantle upwellingrate, which is in turn determined by (1) platespreading rate (R) if sub-ridge mantle upwellingis mostly passive and (2) the component of dy-namic upwelling. The dynamic upwelling is shownto increase with decreasing spreading rate [34],and must also depend on thermal (T) or composi-tional (X) buoyancy of the upwelling mantle rela-tive to its surroundings. Therefore, Mf = f(R,T,X).The extent of mantle melting beneath ocean ridgesis generally thought to be controlled by mantletemperature variation [27^29] and plate spreadingrate variation [13,30], but it must also depend onthe composition of the fertile mantle material.More melt would form at a given mantle temper-ature and for a given amount of decompressionfrom a mantle that contains more enriched com-ponent than a more refractory mantle. Thus,F = f(R,T,X) is also true. Therefore, variations inplate spreading rate, mantle temperature and fer-tile mantle composition are the fundamental var-iables that determine the ultimate melt productionand crustal mass at ocean ridges: Mc = f(R,T,X).

As the plate spreading rate is essentially thesame, V22 mm/yr [35] at both OH-1 and OH-3,

Fig. 4. Variations of representative incompatible elementabundances (Ba, U, Ta and Sr in ppm) and ratios (Nb/La,La/Sm, Zr/Y and Hf/Yb) as a function of sampling depthfor lavas from both OH-1 and OH-3.

EPSL 5771 22-3-01


the e¡ect of spreading rate variation can be ruledout as an in£uential factor, which leaves mantletemperature and source compositional di¡erencesas the possible controls. That is, Mc = f(T,X).

4.1.1. Is the mantle beneath OH-1 necessarilyhotter than the mantle beneath OH-3?

OH-1 lavas are highly enriched in incompatibleelements including heat-producing K, U and Threlative to OH-3 lavas. Thus OH-1 mantle sourceregion could be intrinsically hotter than OH-3mantle source region. This e¡ect certainly exists,but it must be insigni¢cant because of the verylong half-lives, thus very slow heat production,of these heat-producing isotopes compared tothe essentially `zero-age' process beneath oceanridges. Furthermore, heat conduction is e¤cient,and thus it is di¤cult to generate and maintain alarge thermal gradient at the mantle solidus depthon spatial scales as short as a few tens of kilo-meters such as between OH-1 and OH-3.

Detrick et al. [5] argued that the systematicsouthward deepening of the MAR rift valleyand decrease in MBA between the Oceanographerand Hayes transforms are consequences of twoe¡ects. One is increasing distance from the Azoreshotspot to the north, and the other is `a system-atic decrease in the segment length/o¡set ratiofrom north to south between the two transforms'([5], p. 3775). These interpretations are inconsis-tent with the observations. If thermal in£uence ofthe Azores hotspot indeed caused the contrastingdi¡erences between OH-1 and OH-3, then ridgesegments north of the Oceanographer transform(e.g. PO-8, PO-7) would have much greater e¡ect.That is, these segments would have shallower ax-ial depth, greater negative MBA, and moreAzores-like basal compositions, but none of theseis observed [5,17,18]. Fig. 5 shows that OH-1 hasthe greatest negative MBA of all the ridge seg-ments studied at the North MAR, greater thanall ridge segments closer to the Azores [3^5,24,25]. It is simply unlikely that the Azores hot-spot would have greater thermal in£uence on OH-1 than on ridges closer to the hotspot itself.Therefore, the contrast between OH-1 and OH-3cannot be ascribed to Azores' thermal e¡ect.

Fig. 5. Plots of axial variation in MBA (mGal) against (a)ridge segment length, (b) o¡set size and (c) length/o¡set ratioafter [5]. Note that the well-de¢ned signi¢cant inverse corre-lation in (a) and the scattered correlation in (b) point to anintrinsic link between these parameters in their origin. Notealso that the non-correlation in (c) suggests that the originof MBA is independent of `cold-edge' e¡ects at lithosphericlevels, and thus must ultimately result from sublithosphericmantle processes. Open squares are data from the MAR 28^31³N [3] and solid circles are data from the MAR 33^37.5³Nincluding OH-1, OH-2 and OH-3 [5]. The shaded arrow in(c) points to the direction of increasing `cold-edge' e¡ects.

EPSL 5771 22-3-01


Hence, the mantle source region is not hotter be-neath OH-1 than beneath OH-3. It is possible thatthe enriched OH-1 material could have come fromthe Azores plume (i.e. a compositional e¡ect), butthis hypothesis is unsupported isotopically [18].

By using a passive, plate-driven mantle £owmodel, Detrick et al. [5] concluded that the mantleis hotter beneath OH-1 than beneath OH-2 andOH-3 by assuming a southward decrease in thesegment length/o¡set ratio from north to south.This along-axis temperature variation would con-tribute to the observed MBA (the dashed line inFig. 1d). Fig. 5c shows however that the segmentlength/o¡set ratio of OH-1 is the smallest, nothighest. This means that if `cold-edge' e¡ectwere important, OH-1 would be cooled mostnot only because of its low length/o¡set ratio,but also because of the cold and old (V10 Ma)lithosphere across the permanent Oceanographertransform to the north. In contrast, OH-2 andOH-3 are only o¡set by short non-permanent dis-continuities, which are probably not even coldedges. The scatter in Fig. 5c demonstrates thatalong-ridge morphology, topography and MBAare independent of segment length/o¡set ratio,and thus are independent of the `cold-edge' e¡ect.

4.1.2. Fertile mantle compositional control on theextent of mantle melting and crustproduction

The above discussion leaves fertile mantle com-position as the sole likely variable responsible forthe contrasting di¡erences between magmaticallyrobust OH-1 and starved OH-3, i.e. Mc = f(X).Indeed, the huge lava compositional di¡erencesbetween OH-1 and OH-3 (Figs. 2^4) can only be

CFig. 6. a: Solidus temperature of anhydrous peridotites at agiven pressure (e.g. 2 GPa) decreases with increasing total al-kali (Na2O+K2O) contents. HPY, MPY and TQL are modelmantle compositions studied experimentally by [59] andKLB-1 by [60]. The total alkali contents for a mean OH-1source (V0.36 wt% Na2O and V0.09% K2O) and a meanOH-3 source (6 0.17% Na2O and 6 0.01% K2O) are esti-mated by using the average data, mean extents of meltingcalculated in Fig. 7, and polybaric melting models of Niuand Batiza [13,28]. b: This shows the fact that the depth atwhich an ascending mantle begins to melt varies signi¢cantly,depending on its fertility (total alkalis) even if the mantle po-tential temperature is assumed to be the same (e.g. 1350³C).c: The e¡ect of H2O in lowering the solidus temperaturefrom a dry solidus is linearly interpolated from the solidusof wet peridotite with 0.1% H2O by Wyllie [38]. Note thatalthough experimental studies on hydrous peridotite meltingexist (e.g. [38^41,59^62]), no constrained melting experimentson peridotites with 6 0.1% H2O are available. Note alsothat the source water content for OH-1 (V280 ppm) andOH-3 (V70 ppm) is calculated as in Fig. 7.

EPSL 5771 22-3-01


explained by melting of compositionally distinctand heterogeneous mantle sources. That is, OH-1 lavas re£ect a heterogeneous and enriched fertilesource, whereas OH-3 lavas re£ect a heterogene-ous and depleted source. Regardless of detaileddi¡erences in fertile source mineralogy, mantleperidotites enriched in alkalis and other incom-patible elements have lower solidus temperatures,thus begin to melt deeper, melt more, and pro-duce more melts than peridotites depleted of theseelements at a given mantle potential temperatureas demonstrated in Fig. 6a,b. It is also well-estab-lished that H2O, the most abundant volatile inEarth's mantle, behaves like an incompatible ele-ment (e.g. similar to K, La, Ce) during MORBgenesis, and MORB melts enriched in incompat-ible elements always have elevated H2O contents[36,37], re£ecting their similar inheritance in themantle sources. Therefore, OH-1 source (vs. OH-3source) must also have elevated H2O water con-tents, which further reduce the solidus tempera-ture signi¢cantly (Fig. 6c). While further experi-mental re¢nement is needed, existing data indicateconsistently that the e¡ect of 0.05% (500 ppm) ofH2O in mantle peridotite is equivalent to loweringthe solidus temperature in excess of 200³C atV100 km depth [38^41].

The actual water contents in the studied sam-ples are unanalyzed, but it is straightforward toestimate water contents in these lavas using theobserved H2O^Ce relationship characteristic ofNorth MAR MORB (H2O = 259.22UCe+172.14;r = 0.969; in ppm) [37]. Calculations show thatOH-1 lavas have 0.44 þ 0.24 wt% water whereasOH-3 lavas have on average only 0.19 þ 0.09 wt%water (see EPSL Online Background Dataset1). Todetermine precisely the amount of water presentin the OH-1 and OH-3 mantle sources and itse¡ect in enhancing the extent of melting is di¤-cult, but the relative di¡erences between OH-1and OH-3 sources can be estimated satisfactorilyas done for Ce in Fig. 7. Fig. 7a shows that themean OH-1 source is signi¢cantly more enrichedthan the mean OH-3 source with more incompat-ible elements being more enriched than less in-compatible ones as expected. For example, Nbin the OH-1 source is V20 times higher than inthe OH-3 source, and Nd is only 2.5 times higher.

Note the 4-fold H2O enrichment in the OH-1source relative to the OH-3 source. This di¡erencein source H2O abundances further promotes thegreater extent of melting and melt production atOH-1 than at OH-3.

Fig. 7b plots primitive mantle normalized mod-el elemental abundances in the OH-1 (V280 ppmH2O) and OH-3 (V70 ppm) sources. The modelsalso calculate that the average composition ofOH-1 lavas can be produced by FOHÿ1 =V10%melting of the OH-1 source, over twice as much asFOHÿ3 =V4.5% melting for OH-3 lavas from theOH-3 source. The di¡erence in the extent of melt-ing between OH-1 and OH-3 must, however, beintrinsically related to their di¡erences in sourcefertility (i.e. alkali and H2O abundances etc.). As-suming a mantle potential temperature of 1350³C,then the depleted OH-3 mantle source with V70ppm H2O begins to melt at 6 2 GPa (6V60km), whereas the enriched OH-1 source withV280 ppm H2O begins to melt at V3 GPa(V90 km) (Fig. 8). This di¡erence in the initialdepth of melting (Po) is considerable and has im-portant consequences. Clearly, the more fertileOH-1 mantle begins to melt deeper, has greatermelting interval, and thus melts to a greater ex-tent than the more refractory OH-3 mantle. Itis also likely that the more fertile OH-1 mantlewill produce more melt per unit pressure releasethan OH-3 mantle, thus contributing to thegreater extent of melting of the OH-1 mantle.This also explains the `garnet signature' conspic-uous in OH-1 lavas but not so in OH-3 lavas (e.g.Zr/YOHÿ1 = 3.46 þ 1.26sZr/YOHÿ3 = 2.15 þ 0.44 ;[Sm/Yb]N;OHÿ1 = 1.35 þ 0.37 s [Sm/Yb]N;OHÿ3 =1.02 þ 0.12).

Importantly, ine¤cient melt^solid separation atslow-spreading environment leads to signi¢cantmelt-retention in the upwelling and melting man-tle [42,43], which in turn promotes buoyant up-welling and enhances pressure release melting,crustal production, residual mantle depletionand density reduction ^ enhanced along-axis grav-ity low. This e¡ect ultimately leads to focused 3-Dupwelling on ridge segment scales [3,4,44]. Suchmelt-retention enhanced focused upwelling mustbe better developed beneath OH-1 than beneathOH-3 as no melt forms beneath the latter until the

EPSL 5771 22-3-01


rising mantle reaches a much shallower level (Fig.8b). This interpretation is consistent with all theobservations (Figs. 1^7).

4.2. The relationship between MBA and magmaticsegmentation at slow-spreading ridges

There is agreement that along-axis MBA varia-tion within a ridge segment re£ects crustal thick-ness variation ^ more negative MBA associatedwith thickened crust at segment center and weakMBA with thin crust at segment ends. That is,melt formed in the mantle is delivered more tothe segment center than to the segment ends. Itcould be that such along-axis variations in melt

Fig. 7. a: Ratios of average OH-1 lava compositions overaverage OH-3 lava compositions plotted for trace elements inthe order of decreasing incompatibility [32]. H2O* is calcu-lated H2O content (see EPSL Online Background Dataset1).The estimated source ratios are plotted to show that the ob-served lava compositional di¡erences between OH-1 andOH-3 are inherited from their respective sources. b: Primitivemantle [63] normalized trace elemental abundances of aver-age OH-1 and OH-3 lavas and their respective model sourceabundances. The average OH-1 lavas can be produced by10.1 þ 1.7% non-modal batch melting of the OH-1 source,whereas the average OH-3 lavas represent 4.5 þ 1.7% non-modal batch melting of the OH-3 source. The source ratios,source abundances, and the extents of melting are estimatedas follows based on a number of assumptions. Assuming Ba,the most incompatible element among commonly analyzedtrace elements [32], has a bulk distribution coe¤cientDBa = 0, we can plot CBa/Ci against CBa (CBa and Ci are con-tents of Ba and element i in lavas). This gives a linear equa-tion of the form CBa/Ci = SiCBa+Ii (Equation 1) [64,65] foreach element i considered for OH-1 and OH-3 sample suitesrespectively. The slopes, Si = Di

o/Cio (Equation 2), and inter-

cepts, Ii = CBao /Ci

o(13Pi) (Equation 3), are readily obtainedby linear regressions, most of which have rv0.9, statisticallysigni¢cant at s 99% con¢dence levels. Where CBa

o and Cio

are source abundances of Ba and element i ; Dio is the initial

bulk distribution coe¤cient of element i, and Pi is the e¡ec-tive bulk distribution coe¤cient of element i depending onmineral melting modes [64,65]. Assuming Di

o is the same forboth OH-1 and OH-3 (a reasonable approximation for thispurpose, but not true for Pi), then the source ratio

Ci;OHÿ1o /Ci;OHÿ3

o = Si;OHÿ3/Si;OHÿ1 (Equation 4) as plotted in a.The source abundances for OH-3 are estimated fromCi

o = Dio/Si (Equation 5), where Di

o is calculated by assumingmelting in the spinel lherzolite stability ¢eld with initialmodes of 55% olivine, 30% orthopyroxene, 13% clinopyrox-ene, and 2% spinel [66], and by using published mineral^ba-salt partition coe¤cients (see [66] for compilation). The ex-tent of melting for OH-3 is calculated using F = (Ci;OHÿ3

o /Ci3Di

o)/(13Pi) (Equation 6), where Pi is calculated using thesame mineral^basal partition coe¤cient data and the decom-pression melting modes of [13]: 0.466 Cpx+0.681 Opx+0.048Spl = 0.193 Ol+1 Melt. FOHÿ3 = 4.5 þ 1.7% is the average and1c value of calculated F considering all the rare-earth ele-ments in all the OH-3 samples with [La/Sm]N 6 1. The sourceabundances for OH-1 were calculated by combining Equa-tions 3 and 4. Calculating the extent of melting for OH-1 isnot straightforward because Ii values are negative for mostheavy rare-earth elements, suggesting Pi s 1, and garnet mustbe a residual phase holding back heavy rare-earth elements.Thus, the polybaric decompression melting reaction for melt-ing in the spinel lherzolite depth range does not apply. How-ever, by assuming CBa

o = 5.32 ppm (CNbo U8.58, a value as-

sumed to be Ba/Nb ratio in the highly enriched sources), wethen solved F and Pi simultaneously by iteration (combiningEquations 3 and 6) by minimizing the relative standard devi-ation of average F values obtained by considering all therare-earth elements. FOHÿ1 = 10.1 þ 1.7% is such an averageand 1c value. These calculated values in the extent of melt-ing should not be taken as de¢nite, but they do argue per-suasively that the observed di¡erences in morphology, topog-raphy, crustal thickness and MBA between OH-1 and OH-3result from di¡erent extents of melting and overall melt pro-duction ultimately controlled by fertile mantle source compo-sitional di¡erences. Note that the apparent Zr^Hf anomaliesin the calculated sources or source ratios result form thepublished Kd values used, and may have no signi¢cance.6

EPSL 5771 22-3-01


delivery result from along-axis variation in meltproduction. For example, the extent of meltingmay be high in the segment center and low nearthe o¡sets due to subdued upwelling or cold-edgee¡ects [44,45]. It is also possible that there is sub-stantial subcrustal melt migration and focusingtowards segment center and the crustal thicknessvariation results from melt redistribution from thesegment center to segment ends at the crustal level[7,17,45,46]. We can better understand melt deliv-ery processes by analyzing current mantle £owmodels for the MAR.

There are two classes of models : focused man-tle upwelling represented by Lin and co-authors[3,4] and 3-D melt migration by Madge and co-authors [7,46]. The former suggests that magmaticaccretion along the MAR is characterized by fo-cused mantle upwelling and melting at discretecenters initiated by a Rayleigh^Taylor gravita-tional instability developed in the partial meltzone [47^49]. Such focused upwelling develops be-cause of melt-retention enhanced buoyancy withhotter and faster upwelling in the center and cold-er and retarded upwelling at the edges. As a re-sult, more melt is produced by decompression inthe segment center than near the edges, whichexplains, if melt migrates vertically, the thickenedcrust and more negative MBA in the segmentcenters. The 3-D melt migration model requiresuniform mantle upwelling with varying ¢nal depthof melting limited by the `cold' lithospheric lidthat is thin in the segment centers and thickenedbeneath o¡sets. Melt migration is directed by theslopes of the lid towards the segment center, anddelivered to the crust. Melt redistribution in thecrust from the center to the segment ends is in-voked to explain the observed along-axis crustalthickness variation. The fundamental di¡erencebetween these two models is the cause^e¡ect rela-tionship between magmatism and ridge segmenta-tion. In the focused mantle upwelling model, ridgesegmentation is the consequence of focused up-welling at discrete centers. In the 3-D melt migra-tion model however, pre-existing o¡sets determinemelt migration.

We can test these two models against observa-tions. Our preferred model is one that explains thesigni¢cant inverse correlations of axial MBA var-

iation with ridge segment length (Fig. 5), and theobservation that enriched lavas prefer to occupysegment centers (Fig. 3). The test shows that thefocused upwelling model is consistent with theobservations although 3-D melt migration cannotbe entirely ruled out. However, a uni¢ed modelmay emerge from these two models if the e¡ect offertile mantle compositional variation is consid-ered (see Fig. 8).

4.2.1. The origin of the inverse correlation of MBAvariation with segment length

The signi¢cant correlation of axial MBA varia-tion with ridge segment length (Fig. 5a) points toa genetic link between the two parameters. Theinverse correlation of ridge o¡set size (Fig. 5b)and the non-correlation of segment length/o¡setratio with MBA variation (Fig. 5c) indicate thatthe ultimate process responsible for the MBA var-iation is deep and sublithospheric as it is unaf-fected by the `cold-edge' e¡ect at shallow levels.This observation is not in favor of 3-D melt mi-gration model whose action is determined by thecold-edge and cold-lid e¡ects at shallow mantle.3-D melt migration may be invoked to qualita-tively explain the inverse correlation between seg-ment length and axial MBA variation as longersegments would produce more melt which, whenlargely focused at the segment centers, would leadto thicker crust and more negative MBA [24].This action is, however, far inadequate to explainthe in£ated morphology, topography and thick-ened mean crust at long ridge segments likeOH-1.

In contrast, the focused mantle welling modelexplains all these observations. For example, largebuoyancy-driven mantle `thermals' or `diapirs' bytheir size upon development would lead to longerspreading segments because segmentation is deter-mined by the discrete centers of focused mantleupwelling. Longer segments of wide zones of fo-cused upwelling possess greater thermal gradientswith hotter and faster ascent in the center, thusleading to a greater extent of decompression melt-ing and more melt production beneath segmentcenters of longer segments than short segments.Therefore, longer segments (e.g. OH-1) are asso-ciated with thicker crust and more negative MBA

EPSL 5771 22-3-01


than short segments (e.g. OH-3). Focused £owsare 3-D features, and their latitudinal growth orshrink may be accompanied by changes in theirlongitudinal dimension. Therefore, the o¡set sizeof a ridge segment may be genetically associatedwith such changes, thus leading to the inversecorrelation between the o¡set size and axial

MBA variation (Fig. 5b). Such a correlation,however, cannot be perfect because the measuredo¡set size results from the growth/shrink of twoadjacent ridge segments at probably di¡erentrates. A careful examination of the relationshipbetween segment length and o¡set size as a func-tion of time using the existing data [50,51] is

EPSL 5771 22-3-01


needed to quantify the signi¢cance of the correla-tion in Fig. 5b. We emphasize in this context that,as we know, non-transform discontinuities are notpermanent o¡sets, but can migrate as a result ofthe growth of one ridge segment at the expense ofits neighboring segments [33,52,53]. Therefore, itis sublithospheric, not lithospheric, mantle pro-cesses that determine the nature and history of anon-permanent discontinuity, not the oppositeeven though these discontinuities may a¡ect meltmigration/distribution in the crust if any.

4.2.2. Why geochemically enriched lavas prefer tooccupy the center of a segment?

Fig. 3 shows that while depleted lavas occuralong the entire length of a ridge segment, en-riched (OH-1) or less depleted (OH-3) lavas with-in a segment preferentially occur at the segmentcenter. This phenomenon is also observed at someridges in the Paci¢c [54,55], and in particular atmany well-sampled MAR segments [17,33,45],thus is probably a general consequence of mag-matic accretion at the MAR. This observation canbe explained by the model of focused mantle up-welling (Fig. 8b). According to this model, theerupted lavas are delivered vertically from themantle by individual `diapirs' (white arrows).Thus, the along-axis lava geochemical variationre£ects such a variation in the primary mantlemelts. The enriched lavas at the segment centerrequire enriched fertile mantle beneath the center.The thickened crust at the segment center requires

high extents of melting and great melt productionbeneath the center. Both these requirements areconsistent with discrete distributions of enrichedmantle heterogeneities that trigger Rayleigh^Tay-lor type instabilities or `thermals' when they as-cend to their solidus depths. The size and fertility(i.e. the e¡ect of alkali and H2O contents as inFig. 6) of the enriched heterogeneities determinethe initial depth of melting and the size/length ofthe focused upwelling system. A large, and long-lived enriched domain (regardless of its ultimateorigin and history) in the OH-1 mantle has re-sulted in a large focused upwelling system whereasa probably very small amount of enriched materi-al in the OH-3 mantle led only to a small andweak focused upwelling system (Fig. 8b). Hence,the preferred occurrence of enriched or less de-pleted lavas at segment centers is a consequenceof the processes we propose in Fig. 8.

The 3-D melt migration model, however, can-not explain why enriched lavas prefer to occur atsegment center. Subcrustal focusing of melts pro-duced along the entire segment length towardssegment center would e¡ectively homogeneizethese melts before and during delivery to thecrust. As a result, the erupted melts within a ridgesegment would be compositionally uniform, butthis is not the case. Furthermore, melt redistrib-ution from segment center towards o¡sets in thecrust would lead to systematic decrease in lavaeruption temperature (e.g. MgO) as seen inmany segments along the EPR [56,57], but this

Fig. 8. a: Quantitative representation showing the di¡erences in the initial depth of melting (Po), melting depth range (Po^Pf ),thus the extent of melting and melt production between OH-1 and OH-3 because of fertile source compositional di¡erences (alka-lis, H2O and other incompatible elements) even if the mantle potential temperature is the same (e.g. 1350³C). Assuming a mantlepotential temperature of 1350³C, the depleted OH-3 mantle begins to melt at depth 6 2 GPa, but the enriched OH-1 mantlewould begin to melt at a signi¢cantly greater depth, probably V3 GPa Kb. Consequently, OH-1 mantle will have a taller melt-ing interval and melt more than OH-3 mantle. The dry solidi are modi¢ed from [67] considering the data in Fig. 6a,b. The wetsolidi are linearly interpolated from the solidus of wet peridotite with 0.1% H2O by [38]. b: Cartoons showing along-axis di¡eren-ces between OH-1 and OH-3 in the mantle region of upwelling, melting and melt extraction. While mantle upwelling may be uni-form at great depth, focused upwelling begins to develop upon melting because of melt-retention and residue density reductioninduced buoyancy. Such buoyancy-driven £ow, which ascends at elevated rates with enhanced decompression melting and meltproduction, is clearly better developed in the OH-1 mantle, not in the OH-3 mantle where no melt forms until the ascendingmantle reaches a shallower level. In this model, decompression melting and melt production are highest in the center and declinetowards the o¡sets because of the enhanced upwelling rate in the center. Also, we suggest melt migration and delivery to thecrust take vertical paths via small `diapirs' (white arrows) as proposed by [44] without signi¢cant subcrustal melt migration andfocusing. Thus, the observed along-axis crustal thickness variation re£ects along-axis melt production. This model also explainswhy enriched or less depleted lavas prefer to occupy the segment centers (see text for details).6

EPSL 5771 22-3-01


is not observed either. Enriched melts with highH2O contents have lower solidus temperatures.These melts thus have the potential to travel agreater distance towards the o¡sets before solid-i¢cation. Consequently, enriched melts would beexpected to occur near o¡sets, not at segmentcenters, which is again not observed. In fact, lavasfrom many MAR segments are compositionallydiverse and require individual mantle melt batchesdirectly delivered to the crust with their own cool-ing histories [33,44,45].

Alternatively, great melt production and deliv-ery at segment centers may allow the developmentof magma chambers at segment centers, which,through `periodic replenishment, periodic erup-tion, and advanced and continuous fractionation'[58], may contribute to the enrichments in lavaserupted at segment centers. This process is cer-tainly important, if a steady-state magma cham-ber is developed, and can explain the elevatedabundances of incompatible elements, but is inad-equate to explain elevated ratios of incompatibleelements (e.g. [La/Sm]N, [Sm/Yb]N, Nb/Zr, Zr/Y)in lavas erupted at the center of OH-1 withoutinvoking (1) an unusually enriched fertile mantlesource, and (2) deep initial melting evidenced bythe `garnet signatures' (e.g. high [Sm/Yb]N).

4.2.3. Suggestion for a uni¢ed model for magmaticaccretion at the MAR

While focused mantle upwelling model [3,4]with vertical melt delivery to the crust [44] betterexplains both geophysical and geochemical obser-vations than subcrustal melt migration model[7,46], the current focused upwelling model can-not explain why some upwelling systems are de-veloped larger (longer ridge segments) than othersgiven the more or less uniform mantle potentialtemperature in a geographically small area such asthe Oceanographer^Hayes region. It is perhapsthis drawback that led to the rejection of the fo-cused upwelling model because indeed the modelresults [7,46] demonstrate that buoyant £ows havealong-axis wavelengths greater than 150 km, andthus cannot produce the observed size of ridgesegments, which are variably shorter (e.g. 10^100 km). We expect future mantle £ow models

that consider the e¡ect of fertile mantle composi-tional variation (see Figs. 6^8) to succeed in pro-ducing along-axis wavelengths of buoyant £owscomparable to the observed size of ridge segmen-tation at the MAR.

5. Summary

1. The contrasting di¡erences in morphology, to-pography, crustal mass, and MBA betweenOH-1 and OH-3 result from fertile mantlecompositional di¡erence beneath the two ridgesegments.

2. Focused mantle upwelling and melting withvertical melt delivery to the crust best explainsboth geophysical and geochemical observationsat the MAR.

3. The size of focused upwelling is controlled bythe size and fertility of the enriched domainspassively embedded in the sub-ridge mantle.The size and fertility of the enriched domainsmay actually control the initiation and evolu-tion of ridge segments bounded by non-rigiddiscontinuities at slow-spreading ridges.

4. Mantle source compositional variation, as aremantle temperature variation and plate spread-ing rate variation, is a fundamental variablethat determines the physical behavior of theupwelling and melting mantle as well as theextent of mantle melting, crustal productionand MORB chemistry.

Acknowledgements

We thank the captain and crew of the R/V Na-dir and the Nautile group for making the sam-pling program successful. Y.N. acknowledges sup-port from The University of Queensland andARC. D.B. and R.H. thank for support fromIFREMER and French Ministere des A¡airsEtrangeres. R.B. thanks for support from theUniversity of Hawaii and NSF. Discussion withAllegra Hosford and Jian Lin was helpful. Formalreviews by Wolfgang Bach and an unnamed re-viewer are acknowledged.[FA]

EPSL 5771 22-3-01


References

[1] K.C. Macdonald, Mid-ocean ridges: Fine scale tectonic,volcanic, and hydrothermal processes within the plateboundary zone, Ann. Rev. Earth Planet. Sci. 10 (1982)155^190.

[2] J.-C. Sempere, G.M. Purdy, H. Schouten, Segmentationof the Mid-Atlantic Ridge between 24³N and 34³N40PN,Nature 344 (1990) 427^431.

[3] J. Lin, G.M. Purdy, H. Schouten, J.-C. Sempere, C. Zer-vas, Evidence from gravity data for focused magmaticaccretion along the Mid-Atlantic Ridge, Nature 344(1990) 627^632.

[4] J. Lin, J. Phipps Morgan, The spreading rate dependenceof three-dimensional mid-ocean ridge gravity structure,Geophys. Res. Lett. 19 (1992) 13^16.

[5] R.S. Detrick, H.D. Needham, V. Renard, Gravity anoma-lies and crustal thickness variations along the Mid-Atlan-tic Ridge between 33³N and 40³N, J. Geophys. Res. 100(1995) 3767^3787.

[6] M. Tolstoy, A.J. Harding, J.A. Orcutt, Crustal thicknesson the Mid-Atlantic Ridge: Bull's eye gravity anomaliesand focused accretion, Science 262 (1993) 726^729.

[7] L.S. Magde, D.W. Sparks, R.S. Detrick, The relationshipbetween buoyant mantle £ow, melt migration, and gravitybull's eyes at the Mid-Atlantic Ridge between 33³N and35³N, Earth Planet. Sci. Lett. 148 (1997) 59^68.

[8] R.S. Detrick, J. Collins, G. Kent, J. Lin, D. Toomey, A.Barclay, E. Hooft, A. Hosford, S. Hussenoeder, Mid-At-lantic Ridge bull's-eye experiment: A seismic investigationof segment-scale crustal heterogeneity at a slow-spreadingridge, InterRidge News 6 (1997) 27^32.

[9] J.N. Natland, Partial melting of a lithologically heteroge-neous mantle: Inferences from crystallization histories ofmagnesian abyssal tholeiites from the Siqueiros FractureZone, in: A.D. Saunders, M.J. Norry (Eds.), Magmatismin the Ocean Basins, Geological Society Special Publica-tion, 42, 1989, pp. 41^70.

[10] F. Albare©de, How deep do common basaltic magmasform and di¡erentiate? J. Geophys. Res. 97 (1992)10997^11009.

[11] Y. Shen, D.W. Forsyth, Geochemical constraints on ini-tial and ¢nal depth of melting beneath mid-ocean ridges,J. Geophys. Res. 100 (1995) 2211^2237.

[12] Y. Niu, K.D. Collerson, R. Batiza, I. Wendt, M. Rege-lous, The origin of E-type MORB at ridges far from man-tle plumes: The East Paci¢c Rise at 11³20PN, J. Geophys.Res. 104 (1999) 7067^7087.

[13] Y. Niu, Mantle melting and melt extraction processesbeneath ocean ridges: Evidence from abyssal peridotites,J. Petrol. 38 (1997) 1047^1074.

[14] J.-G. Schilling, Azores mantle blob: Rare earth evidence,Earth Planet. Sci. Lett. 25 (1975) 103^115.

[15] J.-G. Schilling, M. Zajac, R. Evans, T. Johnston, W.White, J.D. Devine, R. Kingsley, Petrological and geo-chemical variations along the Mid-Atlantic Ridge from29³N to 73³N, Am. J. Sci. 283 (1983) 510^586.

[16] H.D. Needham, SIGMA Scienti¢c Team, The crest of theMid-Atlantic Ridge between 40³ and 15³N: Very broadswath mapping with the EM 12 echo sounding system,EOS 72 (1991) 470.

[17] C.H. Langmuir, J. Reynolds, H. Bougault, T. Plank, L.Dosso, D. Desonie, E. Gier, Y. Niu, A petrological tra-verse along the Mid-Atlantic Ridge across the Azoreshotspot, J. Conf. Abstr. 1 (1996) 834^835.

[18] L. Dosso, H. Bougault, C. Langmuir, C. Bollinger, O.Bonnier, J. Etoubleau, The age and distribution of mantleheterogeneity along the Mid-Atlantic Ridge (31^41³N),Earth Planet. Sci. Lett. 170 (1999) 269^286.

[19] D. Bideau, R. Hekinian, C. Bollinger, M. Constantin, E.Gracia, C. Guivel, B. Sichler, R. Apprioual, R. Le Gall,Submersible investigation of highly contrasted magmaticactivities recorded on two segments of the Mid-AtlanticRidge near 34³52PN and 33³55PN, InterRidge News 5(1996) 9^14.

[20] D. Bideau, R. Hekinian, B. Sichler, E. Gra©cia, C. Bollin-ger, M. Constantin, C. Guivel, Contrasting volcanic^tec-tonic processes during the past 2 Ma on the Mid-AtlanticRidge: submersible mapping, petrological and magneticresults at lat. 34³52PN and 33³55PN, Mar. Geophys.Res. 20 (1998) 425^458.

[21] E. Gra©cia, D. Bideau, R. Hekinian, Y. Lagabrielle, L.M.Parson, Along-axis magmatic oscillations and exposure ofultrama¢c rocks in a second-order segment of the mid-Atlantic ridge (33³43PN^34³07PN), Geology 25 (1997)1059^1062.

[22] E. Gra©cia, D. Bideau, R. Hekinian, Y. Lagabrielle, De-tailed mapping of two contrasting second-order segmentsof the mid-Atlantic ridge between Oceanographer andHayes fractures zones (33³30PN^35³N), J. Geophys. Res.104 (1999) 22903^22921.

[23] A.H. Barcalay, D.R. Tommey, S.C. Solomon, Seismicstructure and crustal magmatism at the Mid-AtlanticRidge, 35³N, J. Geophys. Res. 103 (1998) 17827^17844.

[24] E.E.E. Hooft, R.S. Detrick, D.R. Toomey, J.A. Collins, J.Lin, Crustal thickness and structure along three contrast-ing spreading segments of the Mid-Atlantic Ridge, 33.5³^35³N, J. Geophys. Res. 105 (2000) 8205^8226.

[25] A. Hosford, J. Lin, R.S. Detrick, Crustal evolution overthe last 2 Ma at the Mid-Atlantic Ridge OH-1 segment,35³N, J. Geophys. Res. (in press).

[26] R. Hekinian, F. Pineau, S. Shilobreeva, D. Bideau, E.Gra©cia, M. Jovoy, Deep seated explosive activity on theMid-Atlantic Ridge near 34³50PN: Magma composition,vascularity and volatile content, J. Volcan. Geotherm.Res. (in press).

[27] E.M. Klein, C.H. Langmuir, Global correlations of oceanridge basalt chemistry with axial depth and crustal thick-ness, J. Geophys. Res. 92 (1987) 8089^8115.

[28] Y. Niu, R. Batiza, An empirical method for calculatingmelt compositions produced beneath mid-ocean ridges:application for axis and o¡-axis (seamounts) melting,J. Geophys. Res. 96 (1991) 21753^21777.

[29] C.H. Langmuir, E.M. Klein, T. Plank, Petrological sys-

EPSL 5771 22-3-01


tematics of mid-ocean ridge basalts: Constraints on meltgeneration beneath ocean ridges, in: J.P. Morgan, D.K.Blackman, J.M. Sinton (Eds.), Mantle Flow and MeltGeneration at Mid-Ocean Ridges, AGU Monograph.,71, 1992, pp. 183^280.

[30] Y. Niu, R. Hekinian, Spreading rate dependence of theextent of mantle melting beneath ocean ridges, Nature 385(1997) 326^329.

[31] J. Cotton, A. Le Dez, M. Bau, M. Caro¡, R.C. Maury, P.Dulski, S. Fourcade, M. Bohn, R. Brousse, Origin ofanomalous rare-earth elements and yttrium enrichmentsin subaerially exposed basalts: Evidence from French Po-lynesia, Chem. Geol. 119 (1995) 115^138.

[32] Y. Niu, R. Batiza, Trace element evidence from sea-mounts for recycled oceanic crust in the eastern equatorialPaci¢c mantle, Earth Planet. Sci. Lett. 148 (1997) 471^484.

[33] P.J. Michael et al., Mantle control of a dynamically evolv-ing spreading center, Earth Planet. Sci. Lett. 121 (1994)451^468.

[34] D.W. Forsyth, Geophysical constraints on mantle £owand melt generation beneath mid-ocean ridges, in: J.P.Morgan, D.K. Blackman, J.M. Sinton (Eds.), MantleFlow and Melt Generation at Mid-Ocean Ridges, AGUMonograph., 71, 1992, pp. 1^66.

[35] C. DeMets, R.G. Gordon, D.F. Argus, S. Stein, Currentplate motions, Geophys. J. Int. 101 (1990) 425^478.

[36] K.E.D. Aggrey, D.W. Muenow, R. Batiza, Volatile abun-dances in basaltic glasses from seamounts £anking theEast Paci¢c Rise at 21³^14³N, Geochim. Cosmochim.Acta 52 (1988) 2115^2119.

[37] P.J. Michael, Regionally distinctive sources of depletedMORB: Evidence from trace elements and H2O, EarthPlanet. Sci. Lett. 131 (1995) 301^320.

[38] P.J. Wyllie, Role of water in magma generation and ini-tiation of diapiric uprise in the mantle, J. Geophys. Res.76 (1971) 1328^1338.

[39] D.H. Green, Experimental melting studies on a modelupper mantle composition at high pressure under water-saturated and under-saturated conditions, Earth Planet.Sci. Lett. 19 (1973) 37^53.

[40] B.O. Mysen, A.L. Boettcher, Melting of a hydrous man-tle: II. Geochemistry of crystals and liquids formed byanatexis of mantle peridotite at high pressures and hightemperatures as a function of controlled activities ofwater, hydrogen, and carbon dioxide, J. Petrol. 16(1975) 549^593.

[41] E. Stolper, S. Newman, The role of water in the petro-genesis of Mariana trough magmas, Earth Planet. Sci.Lett. 121 (1994) 293^325.

[42] D.R. Scott, D.J. Stevenson, A self-consistent model ofmelting, magma migration and buoyancy-driven circula-tion beneath mid-ocean ridges, J. Geophys. Res. 94 (1989)2973^2988.

[43] C. Sotin, E.M. Parmentier, Dynamic consequences ofcompositional and thermal density strati¢cation beneath

spreading centers, Geeophys. Res. Lett. 16 (1989) 835^838.

[44] Y. Niu, R. Batiza, Magmatic processes at the Mid-Atlan-tic ridge V26³S, J. Geophys. Res. 99 (1994) 19719^19740.

[45] J.R. Reynolds, C.H. Langmuir, Petrological systematicsof the mid-Atlantic ridge south of Kane: Implicationsfor ocean crust formation, J. Geophys. Res. 102 (1997)14915^14946.

[46] L.S. Magde, D.W. Sparks, Three-dimensional mantle up-welling, melt generation, and migration beneath seg-mented slow-spreading ridges, J. Geophys. Res. 102(1997) 20571^20583.

[47] J.A. Whitehead, H.J.B. Dick, H. Schouten, A mechanismfor magmatic accretion under spreading centers, Nature312 (1984) 146^147.

[48] H. Schouten, K.D. Klitgord, J.A. Whitehead, Segmenta-tion of mid-ocean ridges, Nature 317 (1985) 225^229.

[49] K. Crane, The spacing of rift axis highs: Dependence ondiapiric processes in the underlying asthenosphere, EarthPlanet. Sci. Lett. 72 (1985) 405^415.

[50] J. Lin, B.E. Tucholke, M.C. Kleinrock, O¡-axis crustalstructure of the Mid-Atlantic Ridge, InterRidge News 6(1997) 23^26.

[51] B.E. Tucholke, J. Lin, M.C. Kleinrock, M.A. Tivey, T.B.Reed, J. Go¡, G. Jaroslow, Segmentation and crustalstructure of the western mid-Atlantic ridge £ank,25³30P^27³10PN and 0^29 Ma, J. Geophys. Res. 102(1997) 10203^10223.

[52] N.R. Grindlay, P.J. Fox, K.C. Macdonald, Second orderridge axis discontinuities in the South Atlantic: Morphol-ogy, structure, and evolution, Mar. Geophys. Res. 13(1991) 21^49.

[53] P. Gente, R.A. Pockalny, C. Durand, C. Deplus, M.Maia, G. Ceuleneer, C. Mevel, M. Cannat, C. Lavarne,Characteristics and evolution of the Mid-Atlantic Ridgebetween 20³N and 24³N during the last 10 million years,Earth Planet. Sci. Lett. 129 (1995) 55^71.

[54] J.L. Karsten, J.R. Delaney, J.M. Rhodes, R.A. Lilias,Spatial and temporal evolution of magmatic systems be-neath the Endeavor segment Juan de Fuca ridge: Tectonicand petrologic constraints, J. Geophys. Res. 95 (1990)19235^19256.

[55] M.R. Per¢t, D.J. Fornari, Geochemical studies of abyssallavas recovered by DSRV Alvin from eastern GalapagosRift, Inca Transform, and Ecuador Rift: 2, Phase chem-istry and crystallization history, J. Geophys. Res. 88(1983) 10530^10550.

[56] R. Batiza, Y. Niu, Petrology and magma chamber pro-cesses at the East Paci¢c Ridge V9³30PN, J. Geophys.Res. 97 (1992) 6779^6797.

[57] M. Regelous, Y. Niu, J.I. Wendt, R. Batiza, A. Greig,K.D. Collerson, An 800 ka record of the geochemistryof magmatism on the East Paci¢c Rise at 10³30PN: In-sights into magma chamber processes beneath a fast-spreading ocean ridge, Earth Planet. Sci. Lett. 168(1999) 45^63.

EPSL 5771 22-3-01


[58] M.J. O'Hara, R.E. Mathews, Geochemical evolution inan advancing, periodically replenished, periodicallytapped, continuously fractionated magma chamber,J. Geol. Soc. Lond. 138 (1981) 237^277.

[59] D.H. Green, T.J. Falloon, Pyrolite: A Ringwood conceptand its current expression, in: I. Jackson (Ed.), TheEarth's Mantle, composition, Structure, and Evolution,Cambridge University Press, 1998, pp. 311^380.

[60] E. Takahashi, Melting of a dry peridotite KLB-1 up to 14GPa: Implications on the origin of peridotitic upper man-tle, J. Geophys. Res. 91 (1986) 9367^9382.

[61] G.A. Gaetani, T.L. Grove, The in£uence of water onmelting of mantle peridotite, Contrib. Mineral. Petrol.131 (1998) 323^346.

[62] K. Hirose, Melting experiments on lherzolite KLB-1under hydrous conditions and generation of high-magne-sian andesitic melts, Geology 25 (1997) 42^44.

[63] S.-s. Sun, W.F. McDonough, Chemical and isotopic sys-tematics of ocean basalt: Implications for mantle compo-sition and processes, in: A.D. Saunders, M.J. Norry(Eds.), Magmatism of the Ocean Basins, Geol. Soc.Lond. Spec. Publ. 42, 1989, pp. 323^345.

[64] D.M. Shaw, Trace element fractionation during anatexis,Geochim. Cosmochim. Acta 34 (1970) 237^243.

[65] A.W. Hofmann, M.D. Feigenson, Case studies on theorigin of basalt: I. Theory and reassessment of Grenadabasalts, Contrib. Mineral. Petrol. 84 (1983) 382^389.

[66] Y. Niu, R. Hekinian, Basaltic liquids and harzburgiticresidues in the Garrett Transform: A case study at fast-spreading ridges, Earth Planet. Sci. Lett. 146 (1997) 243^258.

[67] D. McKenzie, M.J. Bickle, The volume and compositionof melt generation by extension of the lithosphere, J. Pet-rol. 29 (1988) 625^679.

EPSL 5771 22-3-01


mantle compositional control on the extent of mantle...

Documents